During the course of the last year, our team has had a number of conversations about product reliability and Bayesian methods with Eric Maass (Sr. Director, Design/Quality/Manufacturing) and Mark Porter (Sr. Eng. Mgr./Technical Fellow) at Medtronic in Tempe, Arizona. These conversations lead to a session we delivered for Mark’s team on Wed., November 4th (with input on Bayesian Reliability from Chris Gotwalt). Medtronic’s use case is interesting and may be useful for us internally.
Medtronic’s adoption of Bayesian methods came about after a highly damaging 2007 recall of a faulty heart defibrillator wire that caused 13 fatalities, as well as random, non-lethal shocks to some of the 235,000 patients who received the implant. The resulting class action lawsuit ended with a settlement of $268 million, and decline of over $1 billion in the company’s valuation. Eric also described the personal impact it had on the teams involved, with many feeling personally responsible for having harmed their patients.
In the aftermath, one of the teams involved undertook an effort to re-analyze the historical data. They were able to demonstrate to Medtronic’s board of directors that the problem would most likely have been identified prior to product release had Bayesian methods been used from the start of development. We were not able to learn the specifics of the analyses only that they involved Bayesian methods.
Bayesian methods have since been adopted throughout Medtronic for reliability demonstration calculations. Calculations are made using an R script written by statisticians at Medtronic’s Minneapolis site.
The FDA now accepts the use of Bayesian statistics in medical device clinical trials. There is a guidance document published by the FDA. There are also many publicly available papers on the subject, some of which were authored or co-authored by FDA employees. This is another indication of a trend up in the use of Bayesian methods and a potential opportunity for JMP.
Mark Porter’s current project is to explore Bayesian Reliability in the context of Medtronic’s microprocessor manufacturing operations located in Tempe, which was the catalyst for the session delivered on Nov. 4th (see attached project archive).
With the help of Wendy Tseng -SE for the Medtronic sites in the Minneapolis area -the conversation has advanced to reveal strong interest in providing the same presentation to a wider audience, as well as an opportunity to provide formal training on Bayesian Reliability through SAS Education.
We are excited to explore Bayesian Reliability further with Medtronic and look forward to seeing whether our work might have wider application across JMP’s development and sales teams.
